[PATCH 0/7] RFC Memory tagging support

Richard Earnshaw Richard.Earnshaw@foss.arm.com
Mon Jun 15 16:37:18 GMT 2020


On 15/06/2020 16:08, Paul Eggert wrote:
> On 6/15/20 7:40 AM, Richard Earnshaw wrote:
>> 3) Tests that construct a fake internal malloc data structure and then
>> try to perform operations on them.  I haven't looked at these in too
>> much detail, but the first issue is that the fake header is only
>> 8-byte aligned and for MTE to work it requires a 16-byte aligned
>> structure
> 
> Is the problem here that the tests fail quickly due to easy-to-check alignment
> issues, and so no longer test the more-interesting defenses?
> 
> Would it make sense for these tests to align the fake internal data structure to
> 16 bytes using _Alignas? I.e., should the tests be trying to fail due to easy
> alignment issues, or should they be trying to skip the easy alignment issues and
> go on to the harder parts of the tests?
> 

It might.  Of course, there's no (easy) way for the test to fake up the
tag colouring, but that's not a major issue (at least on aarch64)
because tags are ignored on non-tagged memory and static data is just that.

This is something I need to look into some more, but I wanted to get the
review started.

R.


More information about the Libc-alpha mailing list